Why Wall Street Is Fighting Over Anthropic's Staffers' Millions

Wealth managers are cutting fees and opening offices around Silicon Valley to win the right to manage the personal money of Anthropic's employees. The company hasn't gone public yet, and its stock doesn't exist to trade. Yet the pitch for that business is already underway, because a payday is coming that could mint a new class of millionaires in a single afternoon.

That payday is why the rest of Wall Street is scrambling, too.

Anthropic, the startup behind the Claude chatbot, is preparing what could be the largest public offering in history. Bankers have told prospective investors it could raise more than $100 billion and value the company at around $2 trillion. The offer hasn't priced, and marketing is expected to begin around mid-October. But the fight over the employees' money is a reliable window into how large the event is expected to be, because banks don't spend this kind of effort on small outcomes.

The bridge to those numbers is a change in the company's economics. By August, Anthropic was on track to run at more than $65 billion in annualized revenue — up more than seven times its pace at the end of last year. Set that growth against its valuation mark — around $380 billion in February — and you can see the market treating the growth rate itself as the asset. This is the training-to-inference transition the entire AI buildout has been moving toward, in its cleanest form: a lab that spent years buying compute to train frontier models is now converting that investment into inference revenue that paying customers actually fund.

The scramble shows up at every level of the banking industry. Because OpenAI is preparing a rival listing, the banks that dominate tech underwriting have been forced to pick teams, forming separate working groups so Morgan Stanley and Goldman Sachs can run both offerings without leaking between them; JPMorgan is also in the deal. And the battle for a seat started before the IPO: Goldman and JPMorgan are finalizing a $15 billion expansion of Anthropic's pre-IPO credit line — a facility that is partly a way to buy access to the offering table. Last quarter alone, Morgan Stanley's wealth division pulled in more than $74 billion of net new assets from IPOs, much of it by managing the employee equity that SpaceX's listing created. The same playbook is now aimed squarely at the two AI labs.

Here is where the excitement and the business separate.

Anthropic's revenue is real and growing unusually fast, but it runs on enormous purchased-compute commitments. The company is among the largest buyers of Nvidia's chips in the world; its disclosed supply agreements run to hundreds of thousands of GPUs at well over a billion dollars a month. That is the tension under the headline: the growth that justifies a two-trillion-dollar valuation is powered by the largest line item in the model, so whether the acceleration reaches the bottom line — or gets consumed by the compute bill — is the open question, not the given. It is also worth noting that, as the listing approaches, the company is considering requiring rank-and-file employees to sell through preset trading plans, a concession to how much insider scrutiny a megacap debut invites.

Then there is what happened to the last giant listing of the AI era. SpaceX went public in June, raised $85.7 billion at a $1.77 trillion valuation, and has since slumped below its offer price, erasing more than a trillion dollars of value; most of this year's large offerings trade below their listing price. That is not a verdict on Anthropic — the businesses are different — but it is a reminder that a record private valuation and a durable public investment are not the same thing, and that enthusiasm which sets a record can reverse quickly when the marginal buyer stops paying up.

For a retail investor, none of this is a signal to buy the IPO — you cannot buy Anthropic at retail in any normal way, and pre-IPO shares trade only among institutions at wide spreads. The useful reading runs deeper. What the scramble over the staffers' millions confirms is that the demand side of the AI buildout is real in dollar terms: revenue growth like this is precisely what years of spending on chips and data centers were supposed to produce, and the wealth-management industry is betting on it. That is the confirmation half of the story, and it matters for anyone trying to place the AI trade.

The other half is the discipline the wealth managers are not advertising. A two-trillion-dollar valuation is a bet that revenue can both keep growing this fast and eventually outrun a massive compute bill at the same time. That combination is the hardest part of the AI investment case, it is entirely unproven at this company, and it is the thing the hype around the largest IPO ever will be least likely to make obvious.